The SwarmManager class uses the 'isShutdown_' variable to keep track of
whether it has been shut down or not. This variable needs to be reset to
false if we want to restart using a SwarmManager after it was shut down,
which we sometimes do when bootstrapping a conversation.
This patch fixes a failure in
ConversationRequestTest::testBanContactRestartAccount that was
introduced by commit eb4213d3a1.
GitLab: #1002
Change-Id: If5d295cbae4edd9e2292d7464e2eb2939b05eedb
This approach allows client apps that install ringtones (and any other resources) in custom locations to specify that directory at runtime. This will take precedence over the build time data directory supplied for platforms which install to a fixed path.
https: //git.jami.net/savoirfairelinux/jami-client-qt/-/issues/1619
Change-Id: I81616f79196e645a5ad677d6956be6a2ffcd976a
The formats.h and video.h header files are needed by the patch but were
not explicitly included. They are implicitly included via internal.h
in FFmpeg 6.0.1 (the version we currently use in the daemon), but not in
later versions (6.1+). (Thanks to Maxim Cournoyer for bringing this
problem to my attention.)
Change-Id: I99f35fa351f6aeca2091a6ee7c267acec0cb08bb
Else, if a client calls addContact with an invalid uri, it will
create a conversation with an invalid contact and the client will
be in a bugguy state.
git.jami.net/savoirfairelinux/jami-client-android/-/issues/1681
Change-Id: Id6227c45c279c78aac0a191b6ae688ebe0d3d1c4
This allow users to delete files. Moreover, change edited type and
use edited commit type instead.
GitLab: #796
git.jami.net/savoirfairelinux/jami-client-qt/-/issues/1287
Change-Id: Ida04c15cf1570de2f7707ce547be9f4a5e638245
shutdown can cause the needSocketCb_ to pass a null socket. In this
case we can directly return instead of locking
GitLab: #995
Change-Id: I8ad744e494ea0bd3fe15f221a09feeb44e9aec83
Conversation::pull lock writeMtx_ while announcing new commits.
If a member changes occurs, this can lock conv->mtx and
other operations can wait on writeMtx_ too.
GitLab: #994
Change-Id: I605ba6d4acb7b25a591be86bf37e9d549eeba8c0
API changed for ut_conference.
For ut_conversation_call, because the host is not a dummy call,
we must change the hangup in manager.cpp and fix sink creation
for host while attaching.
Change-Id: Icfc296949438e92f7b9b986ed6aa3b463ec2334e
This fix testAddAcceptOfflineThenConnects without reverting
7463a8a266
The idea is in case of a sporadic clone failure (peer offline but
present, socket broken while cloning, etc), we schedule a new
clone operation later.
GitLab: #991
Change-Id: Ifacd6bfe383902bff6cb818e41ff8f5ddd0f6917
If the multiplexed socket is shutdown, this will lock the socket
then convModule and other threads can do the reverse causing a
deadlock
Fix ut_conversation_call sporadic deadlocks
GitLab: #991
Change-Id: I5639b9189b7365cf24abeebf4bb80a89ab3b73ff
+ Unbind call from default layer and let the conference manages
the binding to the default ringbufferpool
+ Avoid to bind on receive, upper layers must bind the audio
GitLab: #959
Change-Id: I639e34281668c4d49ff695fa3e38a1896f567f07
Now a conference can start without any call (this avoid to attach
a call without peer to create a conference).
Path is now clearer for a call creation:
+ If we receive a call, we attach it to the conference
+ Else we do not create any call and just attach the host.
Next step is to be able to attach a host in audio only without weird
tricks and group addParticipant/bindParticipant
GitLab: #953
Change-Id: I13785a5525e041c37fb62c0c9f355e9371f1e4ad
Also, start to fix audio binding (this will be a full patch, but
at least do not attach host)
GitLab: #959
Change-Id: I0aeb9121e2880a5afb4b1ca1a30710d1b60a2e7e
deviceId.toString() will be "00000000000..." if empty, causing
weird message to appears.
GitLab: #991
Change-Id: Ic61a2ece6d0ae3d8b93722a4a3d7b700052865f3
